What do we learn from recorded data in our Cirrus aircraft and how best to use it?

Some recent discussions about Cirrus accidents and flying safely prompt this blog. All involved examining the recorded flight data often found, or not, in Cirrus aircraft.

Recent Cirrus events involving flight data

The fatal aerobatic accident at Boynton Beach, FL, in November 2011 provided the most dramatic and widespread interest in what you can do with recorded flight data from a Cirrus. From data extracted from the wreckage and published in the NTSB public information docket in April, we animated three aileron rolls in the accident aircraft and posted those on YouTube. Two online articles in the aviation press, found here and here, picked up the NTSB probable cause reports released a few weeks ago and linked to this COPA animation, resulting in an additional 20,000 views within a week.

Engine problems plaguing several COPA members in the past few weeks prompted them to post their engine data. Other COPA members helped diagnose anomalies, such as cylinder temperatures, oil pressure indications, probes and connector issues, that saved time and money to get them fixed.

The most recent CAPS save in Cheltenham, UK on 6 June 2013 also involves using recorded flight data to understand how the accident happened. Preliminary access to some recorded data provided some insights into the upset, although the data was not complete and further investigation continues.

Finally, a most mysterious CAPS event reminds us of the limitations of recorded data. The parachute failed to deploy near Addison, TX in mid-May 2013. Unfortunately, that 6-pack Cirrus does not have flight data recording capabilities, making the investigation much more challenging.

Recorded data sources in Cirrus avionics

As the Addison incident reminds us, not all Cirrus aircraft routinely record flight data.

Those Cirrus aircraft with 6-pack avionics, including the Sandel SN3308 EHSI, Garmin 430 or Garmin 430W, and STEC autopilots, do not retain any flight data information.  The ARNAV ICDS 2000 MFD has an optional engine monitoring capability that does record data on a PCMCIA card, but few operators go to the trouble to extract the data files, although it can be done.

With the introduction of the Avidyne Entegra MFD, Cirrus operators began routine engine monitoring. Downloaded data files via Zip drives or USB flash memory drives could be analyzed on a laptop or via the web. Uploading those files to www.cirrusreports.com became a common way for pilots, mechanics and knowledgeable experts on the COPA forums to share access to the data and diagnose problems.

The Avidyne Entegra PFD sets the gold standard for flight data recording in a Cirrus. These log files contain attitude information of how the plane flies through the air at a sample rate of 5 Hertz, or 5 samples per second. Several discrete data elements indicate selections made by the pilot when operating the autopilot, navigation and configuration parameters. Unfortunately, access to these PFD log files requires disassembly of the avionics box. Consequently, it takes an accident investigation to justify the effort and usually involves the NTSB Vehicle Recorder Lab or Avidyne technical support to accomplish.

With the introduction of the Remote Data Module (RDM), somewhat crash hardened and located in the tail of the aircraft, we now had a more robust recording capability that often survived high-energy crashes and fires. Not always, but more often. Unfortunately, the RDM samples flight data at a lower rate of 1 Hertz, or 1 sample per second. Sometimes that slower rate misses abrupt maneuvers that can be distinguished if the Avidyne PFD data is extracted.

The Cirrus Perspective records flight data on user-accessible SD cards. This data stream may be limited depending on how the operator manages the insertion and removal of those SD cards, but the same data is logged in the RDM at the same 1 Hertz rate. Two interesting additional discrete status fields show up in these newer Cirrus models: CAPS handle status as stowed or pulled, which definitely determines if and when an occupant activated the Cirrus parachute system.

The Avidyne R9 system also achieves the gold standard with 5 Hertz records of attitude and position information.

At the moment, I have no knowledge of other after market avionics systems and their capabilities for recording flight data, such as Aspen or Garmin 500/600, etc.

Oh, and don't forget. Some of your portable GPS units log flight tracks. Occasionally, the NTSB Vehicle Recorder Lab gets a portable unit and can extract useful data, but not often because of damage from the crash.

What can we learn?

Given all of this data, what use can we make of it?

Perhaps the most common use is uploading engine data for analysis. COPA members use two common web services: www.cirrusreports.com and more recently www.savvyanalysis.com. By uploading the data files from any of the Cirrus avionics, these web tools display strip charts of engine parameters. The data files can be shared publicly or selectively so experts or volunteers can diagnose common difficulties. We see many examples of simple solutions to intermittent problems that have benefited Cirrus owners, such as identifying weak spark plugs, plugged fuel injectors or mistimed magnetos. Professional analysis can be requested with data uploaded to SavvyAnalysis.

Accident reconstruction based on recorded flight data often provides a visual story that numbers alone fail to tell. We get to see the plane on approach to an airport at night (alternatively, a pseudo daylight view) or descending in a 10-turn spin or those dramatic fatal aerobatics at lower and lower altitudes. With enough data, we can look for things that happened earlier in the flight that might contribute to later problems, such as meandering in marginal weather conditions, or failing to track the final approach course before the upset.

These visualizations and reviews of flight data may happen in training environments. Instructors may review practice or real instrument approaches with clients to identify situations that need improvement. Flight schools may routinely examine recorded flight data files for examples where the student or client flew outside of permitted fight parameters, such as bank angles, airspeeds, engine operation lean of peak, etc. An iPad or iPhone app,CloudAhoy, can track portable GPS data and then visualize your flight in Google Maps or Google Earth. 

In commercial and institutional fleets, similar data files and their reviews become the foundation for aviation Safety Management Systems. So far, I'm unaware of any Cirrus operator going this deep, but the airplane and our web infrastructure permits an operator to routinely gather the data and conduct quality assurance reviews like an SMS. If anyone does this with Cirrus aircraft, look for a university aviation department to be first.

How to create an accident visualization?

Based on the interest generated by the Boynton Beach animation, let me describe the process from data file to video on YouTube.

Obviously, the flight data files provide the key ingredients.

In the Boynton Beach accident, the NTSB docket contains three data files, each of which contained enough of the necessary fields. Unfortunately, we do not have timely access to these files and the NTSB often publishes widely varying extracts. This one was lucky. Public access happens only at the factual report stage, so end up waiting many months or years. Then, we find key data elements missing, such as the location or attitude fields.

In some cases, the NTSB has authorized Avidyne or Cirrus to provide COPA with the raw PFD or RDM files. Those contain more details about attitude and various discrete parameters. Unfortunately, the NTSB rarely shows an interest in more data than just the accident flight, so analyzing prior flights that might show patterns cannot be accomplished. But convincing the NTSB to change it's practices is another topic for later . . .

Given location, altitude and attitude information, we can manipulate the data fields into files suitable for either Google Earth KML files or X-Plane FDR files. (Both file formats have sufficient documentation to make this practical.) Those manipulations often require some clever spreadsheet hacking to get the sequence and units into the right template. Occasionally, with jerky data, interpolation with a bi-cubic spline can achieve more realistic flight animations. Fortunately, mostly it's just math!

Finally, either Google Earth or X-Plane can record a movie file of the flight visualization. With X-Plane, synthetic weather can be specified that approximates the reported weather conditions near the accident location, including precipitation, visibility and cloud decks.

The whole process results in a visual representation of data that looks more familiar to pilots and instructors. So, within a minute or two, the whole complex accident sequence can be presented and then discussed for a long time afterwards.

Some gotchas

However, nothing remains simple.

We noted the lower sample rate of RDM. Sometimes, exciting things happen with the plane over short time intervals and we wish that we had higher sample rates of the Avidyne PFD. Consequently, the animations only depict what was sampled and not what actually happened. But that is hard to remember at times.

Often, we lose data when the avionics power off. The Avidyne MFD only writes to the storage file once a minute even though it samples more often. So, we might lose the last minute of data in the accident sequence. Fortunately, the Avidyne PFD and the RDM both write to storage more often.

Finally, the storage devices may be damaged. The NTSB Vehicle Recorder Lab demonstrates a high degree of skill to recover data from broken boxes, cracked circuit boards, burned chips, and even extracting the silicon chip die and using microscope probes to read the data.

All this in search of data that helps COPA Pilots learn how to fly more safely.